-
Notifications
You must be signed in to change notification settings - Fork 15
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
RequestUrl and ResponseUrl using yarl #45
base: url-page-inputs
Are you sure you want to change the base?
Conversation
|
||
def __eq__(self, other) -> bool: | ||
return str(self._url) == str(other) | ||
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The property methods below can be added mapped dynamically with yarl
's. However, we lose the benefit of defining docstrings within them.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We would also lose API governance, so +1 to manual definition.
However, I wonder if we should define them at all for the initial implementation. We want to make sure we get the API right encoding-wise, and if we expose a part of the Yarl interface already as is, I imagine we are introducing the encoding issue in our implementation, with the caveat of not supporting encoded=True
in __init__
to at least prevent Yarl from messing things up.
Maybe the initial implementation should use a string internally instead, and we can convert it into Yarl later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
A good point about handling the encoding. What do you think about setting encoded=False
by default to prevent yarl from messing things up due to incorrect encoding? be37f39. This would be equivalent to having a str
internally, aside from the "smart" helper methods.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The thing is, Yarl exposes encoded
in __init__
as a workaround for proper encoding handling, which they set off not to implement. But I believe what @kmike has in mind is for us to have a URL class that does proper encoding handling, in which case we should probably not expose encoded
at all (maybe encoding
instead, defaulting to "utf8"
).
I would wait for feedback from @kmike before making more API decisions. I am personally not sure of the best approach here, what parts of w3lib.url
we want to apply and how.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think the goal shouldn't be to implement a general URL class; the goal is to implement URL class useful for web scraping.
If that's hard to use yarl or other library's URL class directly, and we're defining the API anyways, we probably should think about it from the API point of view: what's the API we want, what are the features commonly used in web scraping? After figuring out how we'd like API to look like, we can see what's the best way to implement it - wrap yarl, wrap w3lib, do something else.
Based on our previous discussions, I think a scraping-ready URL class should have:
- a way to manipulate query string: add/remove/get/update query parameters
- some kind of urljoin method, probably via
/
operation - probably - a way to extract the domain name?
- anything else?
In addition to this, there is whole bunch of questions about encoding, normalization, converting URLs to ascii-only encoded strings suitable for downloading, etc. The best API to handle all that might require some thought. I wonder if we can side-step it for now somehow.
At the same time, I'm not sure properties like .scheme are that essential. They're are essential for a general-purpose URL class, but are people who write scraping code commonly parse URLs to get their scheme? We can add such methods and properties for sure, but we can do it later. These methods are probably useful for authors of web scraping frameworks / http clients, but less so for people who write web scraping code.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The property methods below can be added mapped dynamically with yarl's. However, we lose the benefit of defining docstrings within them.
It's also about return types - some yarl.URL methods are going to return yarl.URL objects, while here it would make more sense to return _Url
objects.
web_poet/page_inputs/http.py
Outdated
@@ -18,12 +19,46 @@ | |||
_AnyStrDict = Dict[AnyStr, Union[AnyStr, List[AnyStr], Tuple[AnyStr, ...]]] | |||
|
|||
|
|||
class ResponseURL(str): | |||
class _URL: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We could move this later to w3lib
when we're comfortable in the "smart" URL functionalities that we like.
e9a6ecf
to
bb2a8f2
Compare
bb2a8f2
to
b3c7a0a
Compare
|
||
def __eq__(self, other) -> bool: | ||
return str(self._url) == str(other) | ||
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We would also lose API governance, so +1 to manual definition.
However, I wonder if we should define them at all for the initial implementation. We want to make sure we get the API right encoding-wise, and if we expose a part of the Yarl interface already as is, I imagine we are introducing the encoding issue in our implementation, with the caveat of not supporting encoded=True
in __init__
to at least prevent Yarl from messing things up.
Maybe the initial implementation should use a string internally instead, and we can convert it into Yarl later.
web_poet/page_inputs/http.py
Outdated
def __eq__(self, other) -> bool: | ||
return str(self._url) == str(other) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We should probably make sure we implement this in line with our API expectations.
Mind the following:
>>> example_url_1 = yarl.URL("https://example.com")
>>> example_url_2 = yarl.URL("https://example.com/")
>>> example_url_1 == example_url_2
True
>>> str(example_url_1) == str(example_url_2)
False
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
That's a good point. Although I'm not quite sure if we should have the presence of a trailing /
dictates an equality though.
For example, some sites redirect URLs like "https://example.com/" into "https://example.com". For efficiency, the trailing "/" needs to be stripped to prevent an extra request from being wasted due to redirections.
Because of the different behaviors/expectations, such URLs cannot really be equal as they don't point at the same resource (without the redirections).
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
While that is true for https://example.com/foo
and https://example.com/foo/
, the case of https://example.com
and https://example.com/
is special, since the path of a URL, when not specified, is /
.
>>> URL('https://example.com').path
'/'
>>> URL('https://example.com/').path
'/'
>>> URL('https://example.com/foo').path
'/foo'
>>> URL('https://example.com/foo/').path
'/foo/'
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ahhh thanks for making the distinction clear! Handled this on 292a3b4.
web_poet/page_inputs/http.py
Outdated
def __eq__(self, other) -> bool: | ||
if self._url.path == "/": | ||
if isinstance(other, str): | ||
other = _Url(other) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
If we follow pathlib route, probably there shouldn't be magic like this:
In [1]: import pathlib
In [2]: p = pathlib.Path("/etc")
In [3]: p == "/etc"
Out[3]: False
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
You're right. Initially, I was trying to solve the tiresome casting of str()
everytime, similar to the concern noted in #42 (comment). One thing I realized now would be the problem it presents to end users later on. They might be expecting a "https://example.com"
but misses out that it's actually ResponseUrl("https://example.com")
due to this equality "magic".
Updated this on: 912ba77
web_poet/page_inputs/http.py
Outdated
class ResponseURL(str): | ||
""" URL of the response """ | ||
class _Url: | ||
def __init__(self, url: Union[str, yarl.URL], encoded=True): |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
url: Union[str, yarl.URL]
- this type annotation doesn't allow to create RequestUrl from RequestUrl or from ResponseUrl
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Nice catch! Fixed it in d869024.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I forgot to annotate the other methods in other modules. In any case, we'd be rebasing to #42 if we should proceed with this route. This would include the annotations like 5b0e889#diff-e91005d8dfc7a61689963aafad8b742554243e973f4943ec394cab425bbb61f6R81.
This PR attempts to scrutinize the idea as noted down in #42 (comment).
It's built on top of this PR's branch: